1 Introduction

The affordances of virtual reality (VR) environments have received a lot of attention in recent years. While the consumer market is largely focused on the gaming industry, several companies and researchers have shown interest in exploring the use of VR in education (Checa and Bustillo 2020; Concannon et al. 2019; Hamilton et al. 2021; Jensen and Konradsen 2018; Rospigliosi 2022). As the use of VR grows in education and consumer media, it is important to understand how the significant population of neurodivergent individuals experiences and interacts with this medium. This study explores how a variety of task-relevant and task-irrelevant visual and auditory distractions impact the performance of a neurodiverse population on visual search tasks in VR.

As more VR headsets come on the market, research into the design of VR informal STEM education experiences is growing (e.g., Carrozzino et al. 2010; Fischer 2017; Garcia-Cardona, et al. 2017; Izzo 2017; Koterwas et al. 2018; Yung and Khoo-Lattimore 2019; Zhou et al. 2022). The predicted growth of VR in the near future will allow many new immersive STEM learning opportunities, such as exploring endangered coral reefs, the human body, or the extremes of outer space. VR shows promise to broaden participation in STEM by engaging learners in authentic but otherwise inaccessible learning experiences (Sylaiou et al. 2017; Mikropoulos and Natsis 2011; So and Brush 2008). Further research is needed, however, to understand how this novel technology can be used to motivate meaningful learning (Huang and Liaw 2018), particularly for neurodivergent learners (Lukava et al. 2022). While this study provides a good start to exploring how to design VR environments that are inclusive and effective for neurodivergent learners, much research is still needed to see whether these findings generalize. Future work needs to explore the extent to which individual learner profiles deviate from the patterns observed across the broadly defined neurodivergent population of this study.

1.1 Neurodiversity

The term neurodiversity was first coined by sociologist and autism activist Judy Singer (1998), who argued that the neurology of autistic individuals is a natural part of human variation and not a disorder in need of a cure. The term quickly expanded to include differences in neurology associated with a range of diagnoses, including dyslexia, attention deficit hyperactivity disorder (ADHD), dyscalculia, hyperlexia, dyspraxia, and obsessive–compulsive disorder (OCD), among others (Silberman 2015). The neurodiversity movement argues that disabilities in educational and other contexts are a result of dysfunctional environments and not disordered individuals (Armstrong 2012).

Despite their rise in use, the terms neurodiverse and neurodivergent remain ill-defined, particularly for the purposes of experimental research. These terms are still evolving in their usage and meaning and often focus on self-identification. This is problematic in a research context for a number of reasons. Some who identify as neurodivergent may prefer not to disclose their identity or may identify with a term such as autism or dyslexia rather than neurodiverse/neurodivergent, while others who lack a formal diagnosis but see themselves as neurodivergent may hesitate to use the label without a clinical diagnosis (Santuzzi and Keating 2022). These issues are compounded by unequal access to and use of diagnostic resources as a function of social economic strata and cultural disparities in what behavior is considered problematic or disruptive (Rucklidge 2010; Russell et al. 2016; Shi et al. 2021). Whether to require a medical diagnosis or rely on self-identification remains controversial. On the one hand, advocates of the neurodiversity movement prioritize self-identification and a move away from the medical model (Sarrett 2016). On the other hand, others have criticized self-identification due to the lack of any clear guidelines on who is “in” and who is not (Russell et al. 2020). For the purposes of this study, we take a conservative approach to self-identification. We consider neurodivergent individuals as those who self-report having a diagnosis of ADHD, autism, or a learning disability, as opposed to relying on medical diagnoses. However, the vast majority of neurodivergent individuals in this study were recruited from Landmark College, a campus that exclusively serves neurodivergent learners and includes a formal clinical diagnosis as part of their admission to the college.Footnote 1

1.2 Neurodiversity & VRFootnote 2

VR enables different audiences, including people with disabilities, to access and interact in an engaging and informative way (Walczak et al. 2006). These differences may include visual and hearing impairments, mobility issues, sensory processing difficulties, and/or working memory and attention barriers, and many tend to be co-occurring (Zablotsky et al. 2017; Davis and Kollins 2012). In particular, VR is emerging as a promising tool for children with autism (Rogers 2019; Parsons et al. 2004; Goodwin 2008). However, neurodivergent individuals experience barriers in STEM learning experiences, such as sensory processing differences, distractibility, and social anxiety (Hutson and Hutson 2022; Kulik and Fletcher 2016; Little et al. 2014). Neurodivergent learners may respond differently in VR, depending upon their unique combinations of verbal ability, executive function skills, comorbid profiles, sensory sensitivity, and level of social anxiety (Parsons 2016).

One common feature of neurodiversity involves differences in the subjective experience of stimulus intensity (Gaines et al. 2014; Martin and Wilkins 2022; Zolyomi and Snyder 2021). Responses from neurodivergent individuals to high levels of stimulation can range from simply ignoring the source of the stimulation, all the way to emotional meltdown (McCormicket al. 2016). Moreover, media and stimuli that may seem innocuous to a neurotypical onlooker could be deeply disruptive to those who have sensory sensitivities (Morgan 2019). Finally, there are great individual differences across the abstract label of neurodiversity (as is true of more clinical terms such as autism spectrum disorder and ADHD) and a high degree of heterogeneity in how individuals experience and respond to stimuli (Clince et al. 2016).

VR provides a unique opportunity to manipulate sensory information with a high degree of control. Because of this, VR is an ideal environment for exploring the dynamics of neurodivergence and sensory sensitivities. The immersive nature of VR can lead to a more intense experience where the full visual field can be manipulated. Better understanding the dynamics of VR stimulation for neurodivergent learners will help provide guidance for more inclusive VR experiences.

Neurodivergent learners may experience VR differently than neurotypical learners, depending on their unique profile of verbal use, executive function skills, expression of autism characteristics, and level of social anxiety (Parsons 2016). Prior research has revealed a variety of potential benefits and drawbacks of VR for neurodivergent learners. For example, a number of studies have found VR to be a promising tool for teaching and providing skills training for neurodivergent learners (Bashiri et al. 2017; Kalyvioti and Mikropoulos 2014; Parsons et al. 2004; Wang and Reid 2011). On the other hand, some of these tools have been critiqued for seeking to “normalize” autistic individuals while disregarding their strengths and preferences (Williams and Gilbert 2020). In addition, VR has been associated with a variety of symptoms and effects that may significantly impact neurodivergent learners such as cybersickness, visual fatigue, muscle fatigue, acute stress, and mental overload (Souchet et al. 2022). While many of these negative symptoms impact a broad range of users, their impact on neurodivergent individuals tends to be greater (Schmidt et al. 2021; Souchet et al 2022). As a result, all VR content in this study has been carefully codesigned with neurodivergent individuals to minimize the presence and impact of these issues.

1.3 Distractions & performance

A particular challenge of many neurodivergent learners is avoiding distraction. Distractions can adversely impact the performance of a wide range of cognitive processing profiles, especially those of neurodivergent individuals who may be more attuned to noticing novel stimuli (Vivanti et al. 2017). Furthermore, while some people may say they desire certain types of extraneous information present while completing tasks (e.g., music or video), previous research has shown that individuals’ stated preferences do not always align with their performance (Wilson et al. 2018). The impact of distractions can also differ across modalities (Parmar et al. 2021; Top et al. 2019). Therefore, the current study seeks to explore the impact of a variety of task-relevant and task-irrelevant stimuli and how they may differentially impact neurodiverse learners.

1.4 Visual search task

Visual search tasks have been studied extensively by cognitive psychologists over the past several decades to study how individuals are able to avoid distraction while completing a visual activity (Wolfe 2020). While the tasks vary somewhat across studies, the general paradigm involves having participants search for a target shape or object among a set of distractors. An individual’s search speed is impacted by the similarity of the target and distractor, the relevance of the distractor to the task, and the difficulty of processing the visual stimulus, among other factors. Furthermore, research involving neurodivergent individuals indicates differences in their performance on visual search tasks (O’Riordan et al. 2001; Seernani et al. 2021).

Visual search tasks can take on many forms. For example, Neisser (1964) has used displays of letters, words, and images to investigate information and visual processing. Visual search tasks typical involve searching a visual display for one or more targets. Responses may include pressing a button when the target is found, responding whether or not a target is present, identifying the number of targets present, or tagging the location of the target in some way. Stimuli can include single targets, multiple targets, a conjunction of features, or natural scenes among others (see Fig. 1 for some examples). In this study, participants will look for multiple target images arranged within a 3-D space and will use their controllers to tag targets when they are located (see Sect. 3 for more details).

Fig. 1
figure 1

a Single target search, b muti-target search, c conjunction search, d natural scene search

Full size image

As a well-studied cognitive processing task, the visual search paradigm provides an opportunity to explore the impact of distractions on an individuals’ performance. The paradigm allows for the inclusion of a variety of relevant and irrelevant distractions through multiple modalities. This study manipulated the number of distractors (task relevant), lighting conditions (task irrelevant, same modality), and background sound (task irrelevant, other modality) during the visual search task and looked at the impact on participants’ performance.

1.5 Background sound

For some neurodivergent individuals, background sounds or ambient noise can serve as a comforting or regulatory process; for others, however, it can be a distractor, sometimes stopping efforts toward productivity. For example, when the background sound takes on the form of white noise—sounds that occur when a large variety of audible frequencies played at a constant intensity—individuals with ADHD often exhibit improved speed of word recognition, speech recognition, word recall, and attention (Pickens et al. 2019). Other studies have found that background noise can be a challenge for some neurodivergent learners, particularly those who are autistic. In a classroom study (Kanakri et al. 2017), teachers found noise control to be an important issue for neurodivergent learners, with many of their students opting for the use of hearing protection. Given this complexity, it is unclear how different levels of background noise might impact neurodivergent learners, particularly those who face both attention barriers and sensory sensitivities.

1.6 Codesign

The study reported here is part of a larger NSF funded project (DRL-2005447) aimed at developing a STEM VR informal learning experience that is accessible to neurodivergent learners. The work on this research team is guided by the ethos, “nothing about us without us” (Charlton 1998). As such, all aspects of the research are guided by and incorporate the perspectives and participation of the neurodivergent community. This includes having neurodivergent members on the research team, as well as working closely with neurodivergent codesigners when designing research studies and interventions.

Neurodivergent individuals participated as core members of this project’s design team, constituting more than half the team members and included both full time team members and student interns. Codesigners not only informed the design of this project but also helped shape every aspect of the work from initial idea generation, to prototyping, feedback elicitation, design research cycles, to final product (Dahlstrom-Hakki et al. 2021; Edwards et al. 2022). The materials and content designed for this study were based on initial ideas and stimuli generated by the codesign team (see Fig. 2). Codesigners piloted the study to ensure accessibility and comfort.

Fig. 2
figure 2

Codesign sketches and ideas on left, developed stimuli on right

Full size image

2 Norming data

Prior to conducting the study, norming data were collected to help guide the development of the study materials and design. Data were collected from a total of 110 participants recruited online through the project team’s research networks. Participants were presented with a forced-choice task where they were asked to express their preference for one image of a pair. Participants saw a total of 694 image pairs presented randomly as trials. Each trial pair differed in either brightness, hue, or amount of visual clutter in the scene. Pairs appeared in both possible iterations (each image appearing once on the left side and once on the right side). Participants typically took 20–30 min to go through the 694 trials and were provided with a $25 gift card for their participation.

This norming portion of the study was conducted online using the Pavlovia platform (https://pavlovia.org/) and was preceded by a demographic survey conducted using NoviSurvey. Data were analyzed for preference and consistency of preference, with the latter being assessed based on whether the same preference was expressed when an image pair was presented in both possible iterations. Participants with consistency below 60% were excluded from further analysis, as were three participants who opted not to share their neurodiversity status, leaving a total of 76 participants. In terms of gender identity, 42 participants identified as male, 32 identified as female, and five identified as nonbinary/other. Racially, 52 identified as White, seven as African American, three as Asian American, eight as Mixed Race/other, and five declined to respond. Ages ranged from 13 to 53 years with an average of 21.4. Of these participants, 40 self-reported being neurotypical and 36 self-reported being neurodivergent. Of the neurodivergent participants, 15 reported a diagnosis of ADHD, 10 of autism, and 11 of both. Due to the small group sizes and large overlap, analyses did not further break down the neurodivergent group into smaller subcategories. Analysis included all trials that showed a consistent preference and were not below 300 ms in length to exclude preferences that may have been randomly selected. The 300 ms cutoff was based on prior modeling involving perceptual forced-choice tasks (Huber and Cousineau 2003) that indicated that reaction time distribution onsets were around or shortly after the 300 ms mark.

2.1 Norming results

Preferences for brightness, hue, and clutter on the norming task were analyzed separately, as only one variable was manipulated in each image pair. Brightness trials ranged on a scale from 0, which is almost total darkness, to 10, which is bright white, with equal increments between those two extremes. A generalized linear mixed-effects model (GLMEM) was used to fit participant preference on a trial-by-trial basis using a binomial distribution with 0, indicating preference for the left image and 1 preference for the right. Participant is included as a random variable in the model with brightness level and neurodiversity status as factors (see Table 1). Note the random variable variance approached zero, indicating there was little variability across participants once the other variables were accounted for.

Table 1 Results of GLMEM regression analysis for image preference in brightness trials for consistent trials only
Full size table

As can be expected, both neurodivergent and neurotypical participants had the least preference for brightness levels at the extremes, with an optimal preference for both groups at around the 6–7 brightness level. While the overall pattern of preference was similar across both groups, neurodivergent participants’ preferences were spread more broadly across the range of brightness levels.

Clutter images ranged from level 0, indicating a scene with no objects in it, to level 7, indicating a scene heavily cluttered with objects, with the number of objects added in equal increments across levels. A similar GLMEM model was run for these data and indicated a general preference for more clutter in scenes increasing to level 3 and then plateauing (see Table 2). The data however indicate that neurodivergent participants tended to have a greater preference for empty scenes and ones with more clutter. This effect may have important implications because high visual clutter has been found to negatively impact student performance in the classroom (Godwin et al. 2022).

Table 2 Results of GLMEM regression analysis for image preference in clutter trials for consistent trials only
Full size table

Finally, hue varied across several colors. An initial grayscale image (used as the default comparison variable in the model) was changed to one of six other hues: Blue, Cyan, Yellow, Green, Magenta, and Red. Once again, the overall pattern of preference was similar across neurodivergent and neurotypical participants with a general preference for cooler (Blue/Cyan) over warmer (Magenta/Red) colors (see Table 3).

Table 3 Results of GLMEM regression analysis for image preference in hue trials for consistent trials only
Full size table

This norming data provided guidance for participant preferences for scenes in a 2-D environment and indicated that overall, the pattern of preference is similar across neurotypical and neurodivergent individuals with a slightly broader range of preference for neurodivergent learners. In terms of design considerations based on these norming data, to maximize user comfort and preference, designers should consider using cool color palettes, implementing good lighting, and avoiding empty or low clutter scenes. The norming data, however, do not address two key questions motivating the below study: Is student performance and preference aligned, particularly with respect to clutter?; and do these findings apply to a VR experience? The study below attempts to address these questions.

3 Methods

3.1 Participants

Data in this study were collected from 52 volunteers at two locations, the Museum of Science in Boston, MA and Landmark College in Putney, VT. Participants volunteered to take part in this study by signing informed consent forms and were offered $25 gift cards for participating in the data collection sessions outlined in the procedure section below. The majority of neurodivergent participants were recruited at Landmark College, a college that exclusively serves neurodivergent students.

In terms of gender identity, 20 participants identified as male, 28 identified as female, two identified as genderfluid, one as nonbinary, and one declined to respond. Racially, 38 identified as White, two identified as Hispanic/Latino, two as African American, four as Asian American, three as Mixed Race, and three declined to respond. In terms of neurodiversity status based on self-reporting, nine reported ADHD, six reported autism, four reported a learning disability (LD) and ADHD, two reported LD and autism, six reported ADHD and autism, and 25 identified as neurotypical. Ages ranged from 13 to 47 years with an average of 24.9.

3.2 Materials

Data collection during each session consisted of a short survey followed by a VR data collection session. The survey was delivered through a secure account on the NoviSurvey platform and included an informed consent form (or in the case of minors, parental consent and participant assent forms) and participant demographics. The VR experience consisted of a visual search task delivered using a VIVE Focus 3 headset hardwired to a high-end laptop running an NVidia RTX 3070 video card to ensure no lag during data collection. VIVE Focus 3 controllers were used by participants to interact with the space and tag targets. The project team had two identical VIVE Focus 3 and laptop setups for data collection with all hardware settings kept constant across data collection sessions.

Participants were instructed to search for five target alien creatures among a number of distractor aliens that only differed on one small feature (see Fig. 3, the image on the left has an additional bright end on one of the tentacles). Participants had to visually focus each image to tell which were the targets. Participants were placed on a platform within a half-dome space (see Fig. 4), and the targets and distractors could appear in one of 84 positions arranged within this half-dome space, thereby requiring participants to look around using their headsets (see Fig. 5). Note that, targets and distractor positions were slightly displaced at each location to avoid having uniform lines of objects within the space. Each trial always had five targets, but the number of distractors varied across trials. In addition, the lighting conditions and background sounds were varied to see how they would impact the performance of neurotypical and neurodivergent learners in this experience. This provided for the manipulation of three variables: task-relevant noise (number of distractors), modality relevant noise (lighting conditions), and task-irrelevant noise (background sound).

Fig. 3
figure 3

Target and distractor images used for visual search task

Full size image
Fig. 4
figure 4

3-D space that participants are placed in

Full size image
Fig. 5
figure 5

Distribution of possible distractor locations

Full size image

3.3 Design

Each participant saw a total of 30 visual search trials. The first three trials were practice trials to familiarize participants with VR and the task. The remaining 27 trials were blocked combinations of trials that varied the number of distractors, the lighting condition, and the background sound. There were three levels of each variable leading to a 3 × 3 × 3 within-subjects design. Number of distractors had three levels: 5, 15, and 35; lighting condition had three levels: low, optimal, and high brightness; and background sound was an industrial machine noise and had three levels: none, low volume, high volume. Lighting levels were based on the aforementioned norming data reported on in Sect. 2. The background sound was selected based on feedback from our neurodivergent codesigners, who rated the noise as annoying but not intolerable. Volume levels were set based on a maximal tolerable volume and a midpoint between the high volume and no noise.

3.4 Procedure

Participants typically took about five minutes to complete the demographic survey and informed consent form. Participants were then given a short tutorial on how to put on the headset, adjust the settings for a sharp image, and use the controllers. The experimenter worked with each participant to ensure the headset was on, secure, and comfortable, and the image was clear. Once the experiment began, participants completed three practice trials, during which they received feedback and support from the experimenter. Participated were instructed to look around using the headset to see all possible stimuli and to use either controller to aim at and tag a stimulus by pulling the trigger button on the controller. Participants saw a line extended from their virtual controller to the target to help them aim at their intended target. To tag a target participants pointed at it and pulled the controller trigger button (see https://youtu.be/adskt-Z3Whs?si=ltAOfGIiHsJPyjxj for a demonstration video of the feed from the headset). During the experimental trials, the experimenter was quiet and the participants had been informed to wait until between trials if they had any questions or needed to take a break. It typically took participants about 20 min to complete the experimental trials. At the end of each data collection session, the experimenter helped the participant remove the headset, debriefed each participant, and disinfected the headset and controllers.

3.5 Analysis

Data were analyzed using linear mixed-effects models (LMEM) to predict performance on each individual trial using the lme4 package in R (Bates et al. 2015). The models predicted two dependent variables: response time (modeled using a normal distribution) and number of non-hits (modeled using a Poisson distribution). The response time measure was the total time from stimulus onset until the final target was tagged and gave a measure of speed. The number of non-hits was any tag that was not on a target or on a target that had already been tagged. This gave a measure of accuracy. The models included individual participants and trials as random factors, and neurodiversity status, number of distractors, lighting conditions, and background sound as independent factors. We modeled the main effect of each of the three experimental manipulations, as well as their interaction with neurodiversity status.

4 Results

The results of the mixed-effects models are detailed in Tables 4 and 5. The data show that trials with more distractors tended to take longer to complete and were less accurate, which is not surprising and is consistent with prior work involving visual search tasks. Likewise, when a trial’s lighting condition was high or low, participants tended to be slower and less accurate than when the lighting condition was optimal. Surprisingly, however, the background sound distractors did not seem to have a main effect at either the low- or high-volume levels.

Table 4 Results of GLMEM regression analysis for participant accuracy (lower is better) on VR visual search task
Full size table
Table 5 Results of LMEM regression analysis for participant speed (lower is better) on VR visual search task
Full size table

Looking at the interaction effects of these factors with neurodiversity, we see some significant patterns. While the main effect of neurodiversity is not significant in either speed or accuracy, there were significant interactions with the manipulated factors. The increase in the number of distractors has a significantly greater negative impact on neurodivergent participants, as they tended to be both slower and less accurate as the number of distractors increased. Neurodivergent participants also tended to be slightly worse and less accurate in non-optimal lighting conditions than their neurotypical peers, but this difference only reached statistical significance on the measure of their speed on high brightness trials. Finally, and in contrast to the a priori predictions, the data indicate neurodivergent learners perform better, in terms of both speed and accuracy, on the visual search task in the loud background sound condition than they do in the no-background sound condition, whereas neurotypical participants show the opposite, predicted pattern of performance (see Figs. 6 and 7).

Fig. 6
figure 6

Participant accuracy across background sound conditions

Full size image
Fig. 7
figure 7

Participant speed across background sound conditions

Full size image

Qualitative data were gathered from participants, who were asked to provide their feedback and impressions at the end of each session during the debrief. Most participants indicated they enjoyed the experience and the interaction with the VR environment. Many indicated the dark and bright lighting conditions were challenging. Many also found the background sound jarring and annoying at first but indicated that once they acclimated to it, they could tune it out as white noise.

5 Discussion

The goal of this work was to better understand how distinct environmental factors in a VR experience differentially impact neurodivergent and neurotypical participants. In particular, the study aimed to assess how visual and auditory environmental manipulations impact participant performance on a visual search task. The results indicate that there is a complex interplay between task, environmental factors, and self-reported neurodivergence.

The finding that lighting conditions and number of distractors had a significant impact on all participants falls in line with prior expectations. In addition, the larger impact of distractors on participant performance is in line with other visual search studies involving neurodiverse populations (Canu et al. 2021). What was quite unexpected was the differential impact of background sound on the performance of neurodivergent and neurotypical participants. While neurotypical performance both in terms of speed and accuracy drops with louder background sound, the opposite is true of the neurodivergent participants.

It is important to note, however, that the loud background sound condition merely levels the playing field between neurotypical and neurodivergent participants and does not actually improve the performance of neurodivergent participants beyond their neurotypical peers. The background sound has a negative impact on the performance of neurotypical participants. It is likely that this same impact exists for neurodivergent participants and provides a common floor for performance across the two populations. However, the noise seems to also provides a simultaneous benefit (through a non-overlapping mechanism) to neurodivergent learners that allows them to improve their performance as compared to other conditions.

The exact mechanism for the benefit experienced as a result of the loud background sound is unclear. One possibility is that it acts as white noise that blocks out other external or internal distractors and helps participants focus on the visual search task. This would be in line with studies that have found improved performance for students facing attention barriers. For example, some have found that for children diagnosed with ADHD, the presence of white noise reduced omission rates on a No/Go task (Baijot et al. 2016). Alternatively, the sounds could provide a level of stimulation that helps neurodivergent participants remain alert and on task. This would be in line with studies that have found differential impacts for noise across children with different attention profiles (Söderlund et al. 2010). There is also the possibility that both and/or other mechanisms are at work here. Further research is needed to determine how and why sound improves neurodivergent participants’ performance, whether it is true of other sensory modalities, and whether certain types of sound or other stimulation could provide greater benefit or improved performance without negatively impacting any of the participants.

This study has several limitations worth bearing in mind when considering the impacts of these findings. First, it is important to note that neurodiversity status and other demographics are based on participant self-reporting. Future research should focus on specific aspects of cognitive processing and screen based on those aspects to better characterize differences across the population. In addition, it is possible that research designs using different stimuli may elicit different findings. This seems unlikely in terms of the targets and distractors, as well as the lighting conditions, since this study’s findings are in line with similar work in the field. It is possible, however, that different types and volumes of sound or other stimuli may elicit different impacts on performance. Future work in this area should explore how a range of different types of background stimuli may impact performance.

6 Conclusions and implications

These results carry important implications for designers and developers creating content for neurodivergent participants in VR environments. Design is often guided by participant feedback and preference, and while this is valuable, it does not always translate into superior or improved performance. The background sound in this study was specifically selected to be at the edge of what participants would find tolerable, and indeed, many found it jarring, especially at first. However, the fact that it led to an improvement in neurodivergent participant performance carries important implications for testing the impact of design elements on performance, in addition to collecting feedback on user preference. In prior work by the authors (Dahlstrom-Hakki and Alstad 2019), a similar discrepancy was observed between students’ stated preferences and performance. Understanding these differences may be particularly important for VR educational interventions, where improvements in learning and performance are especially critical and may need to be balanced against elements of the design purely focused on participant comfort or enjoyment.